The field of biomedical research is witnessing significant advancements in the development of Retrieval-Augmented Generation (RAG) systems. These systems aim to improve the accuracy and reliability of large language models (LLMs) in generating responses to complex biomedical questions. Recent studies have focused on enhancing the robustness of RAG systems to adversarial evidence, improving their ability to abstain from answering questions when uncertain, and developing more effective methods for retrieving relevant technical documents. Notably, the integration of energy-based models and Rescorla-Wagner steering has shown promise in improving the reliability and trustworthiness of LLMs in real-world applications. Furthermore, the use of ensemble methods and zero-shot question answering has demonstrated state-of-the-art performance in biomedical question answering tasks. Overall, these advancements have the potential to significantly improve the accessibility and comprehension of biomedical information, paving the way for more accurate and reliable public health communication. Noteworthy papers include: Speaking at the Right Level, which proposes a Controlled-Literacy framework for generating tailored counterspeech adapted to different health literacy levels. Trusted Uncertainty in Large Language Models presents UniCR, a unified framework for confidence calibration and risk-controlled refusal, which has shown consistent improvements in calibration metrics and lower area under the risk-coverage curve.